PS02 Biomarker Analysis. Discussant. Sue-Jane Wang, Ph.D.

Transcription

1 PS02 Biomarker Analysis Discussant Sue-Jane Wang, Ph.D. Associate Director, Adaptive Design, Pharmacogenomics/Phrmacogenetics Office of Biostatistics, Office of Translational Sciences, CDER CDER Statistics Delegate, Inter-center Pharmacogenomics Review Group Presented at FDA/Industry Workshop, September 29, 2006

2 Presentations Opportunity and Challenges in Utilizing Biomarkers for Drug Development Mark Chang, Millenium New Paradigms for clinical Drug Development in the Genomic Era Richard Simon, NCI/NIH Wang SJ, FDA/Industry Biomarker

3 Themes Terminology Some FDA experience Biomarkers: Validation or not? Correlation vs. Prediction Enrichment Strategy Personalized Medicine: new paradigm of statistics? Adaptive design using biomarker Components of Variability Wang SJ, FDA/Industry Biomarker

4 Terminology A characteristic recognized as an indicator Established performance characteristics Single Biomarker Composite Biomarker Wang SJ, FDA/Industry Biomarker

5 Genomic Drug Trial (GDT)* Clinical trials employing (high throughput) genomic technology to identify molecular signals including transcription, SNP or proteomic profiling in complex biological mixtures for use as genomic composite biomarkers (GCB) of disease, of drug exposure/drug disposition, or of drug response including efficacy and toxicity * Wang SJ, Proceedings of the Biopharmaceutical Section, American Statistical Association, 2004 Wang SJ, FDA/Industry Biomarker

6 Experiences in RCT Using Single Genomic Biomarker Her2/Neu (Herceptin( Herceptin)* EGFR (Tarceva( Tarceva*, Iressa) * None of these biomarker subgroups were prospectively planned to evaluate treatment effect for targeted therapy during drug approval cycle Wang SJ, FDA/Industry Biomarker

7 If Tarceva EGFR+ subgroup is pre-specified with appropriate alpha-allocation 0 1 **EGRF (15%) 1.01 ( ) **EGRF + (17%) 0.65 ( ) **EGRF? (68%) HR (95% CI) 0.76 ( ) Overall 0.73 ( ) In favor of Tarceva * Wang et al. 2006, The Pharmacogenomics Journal ** data extracted from Tarceva Package Insert In favor of Placebo Overall p<0.002 Wang SJ, FDA/Industry Biomarker

8 Validation of Genomic (Composite) Biomarker? These biomarkers are used to select responsive patient population Genomic biomarker Surrogate biomarker It is not about validation of surrogate biomarker (surrogate endpoint) Rather, it is about whether the assay can be reproducibly measurable on the biomarker status Wang SJ, FDA/Industry Biomarker

9 Correlation vs. Prediction Correlation does not speak to causal effect Correlation Prediction, especially in exploratory studies OR is often used to describe if a biomarker is predictive of clinical outcome, e.g., in nested ca-co co study A loosely defined criteria for predictivity Wang SJ, FDA/Industry Biomarker

10 Enrichment (Design #1) Goal: increase power of the study Example: GCB+ T > C } Enrichment GCB T = C Design Severe T >> C } Enrichment Mild T > C Design Wang SJ, FDA/Industry Biomarker

11 Enrichment (Design #1) Often perceived to provide predictive ability of the therapeutic response Concerns Generalizability issue Risk:Benefit How many less sensitive patients? Wang SJ, FDA/Industry Biomarker

12 New Paradigm for PM? Emphasize safety and reduce statistical requirement for efficacy? Are we going back to pre-desi era? Criteria to assess B:R balance should not be changed because of the desire to realize PM lack of effect, undesirable characteristics for PM Prudent in research for targeted therapy, a way closer to PM Oncology needs predictive biomarkers (Simon, Wang, TPJ 2006) Wang SJ, FDA/Industry Biomarker

13 Use of GCB profiling in Genomics Drug Trials Genomic Drug Trial Baseline disease phenotype data, etc. Conventional Clinical Trial Therapeutic phenotype data (clinical outcome) Baseline genomic sample (e.g., tissue, specimen, blood, etc.) Is an established GCB classifier or a PG diagnostic assay available at study baseline? Yes No Stratified randomization is an option in addition to a pre-specified statistical analysis plan Prospective planning of a statistical analysis strategy to also studying the therapeutic effect in the GCB subgroup(s) Wang, Temple, O Neill, 2006 Wang SJ, FDA/Industry Biomarker

14 Classifier from high throughput Identify a patient population that benefits from therapeutic Q: how to identify such a patient population Conventional approach Targeted approach Basic concept of demonstrated drug effect remains Assay sufficient? Wang SJ, FDA/Industry Biomarker

15 Figure 5a. Subgroup power for δs/δ = 1, 2 or 3, given α = 0.025, α1 = 0.02, 1-β1= β Wang, Hung, JSM proceeding 2005 sample size ratio (M/N) ds/d=1 ds/d=2 ds/d=3 Wang SJ, FDA/Industry Biomarker

16 Adaptive Design using Biomarker Use of internal data Use of external data Wang SJ, FDA/Industry Biomarker

17 Application of diagnostics to drug studies The more pressing elements are the reproducibility, reliability and precision of the diagnostic or imaging test that are of primary concern for evaluating the test s s predictability in the targeted patient population * Imaging Biomarker for Clinical Prediction, Mills, Wang, Farrell, Pazdur, Woodcock, 2006 Wang SJ, FDA/Industry Biomarker

18 Disease Variability Variable Therapeutic Responses GCV (%) Genetic Component of Variability (GCV) AUC(0-inf) - Reference Drug High (> 80) Intermediate (60-79) Low (<60) b b a 14b c 32a b Drug Number -150 * Lesko, Patnaik, Wang (2003): The Pharmacogenetic Component of Variability in Human Drug Exposure, Sigma/FDA Science Forum, DC, MD Wang SJ, FDA/Industry Biomarker